Researchers create voice-altering technology to disrupt phone scam schemes

A team of researchers from Israel and India has developed a new tool known as ASRJam to combat automated voice call scams. This technology aims to disrupt automatic speech recognition systems used by scammers to carry out fraudulent schemes over the phone.

Voice-based phishing, also referred to as “vishing,” involves using text-to-speech technology to create realistic-sounding voices that lure victims into providing sensitive information or taking certain actions. Criminals attempt to engage victims in conversations in the hopes of obtaining valuable data by converting vocal responses into text for analysis.

According to CrowdStrike’s 2025 Global Threat Report, vishing attacks rose by 442 percent between the first and second half of 2024. The US Federal Communications Commission declared the use of AI-generated voices in phone calls illegal during the first half of that year. Crystal Morin, a cybersecurity strategist at Sysdig, warned that as AI models advance, voice-based phishing is becoming increasingly challenging to detect.

Researchers Freddie Grabovski, Gilad Gressel, and Yisroel Mirsky have introduced ASRJam, which employs an algorithm called EchoGuard to inflict subtle audio perturbations on human speech. By disrupting the performance of automatic speech recognition systems, ASRJam aims to thwart vishing attempts. This defense mechanism focuses on confusing AI systems while ensuring human comprehension remains intact.

The team’s approach is outlined in a paper titled “ASRJam: Human-Friendly AI Speech Jamming to Prevent Automated Phone Scams.” They emphasize that ASR technology represents the weakest link in scammers’ setups and believe their proactive defense method is the first of its kind to be practical for deployment.

EchoGuard, the core of ASRJam, operates in real-time on end-user devices and remains hidden from attackers, making it harder to bypass. The algorithm’s adaptable nature allows it to challenge various AI models effectively, without requiring sample ASR output to create an audio perturbation capable of disrupting the ASR model. EchoGuard uses reverberation, microphone oscillation, and transient acoustic attenuation to modify voice characteristics without compromising clarity and comprehensibility.

To validate the efficacy of ASRJam, the researchers compared it with other ASR jamming techniques using public datasets and six ASR models. EchoGuard consistently outperformed baseline jammers across all tests, with minor exceptions for specific ASR systems. The researchers maintained that their method’s performance against general ASR systems is acceptable, as it excels in real-world deployment scenarios.

By showcasing the superiority of ASRJam over alternative jamming methods and emphasizing its effectiveness in combatting automated voice scams, these researchers have contributed a valuable tool to the ongoing fight against phone vishing schemes. Their work highlights the importance of developing innovative solutions to mitigate cybersecurity threats posed by evolving technologies in the digital landscape.